SWEDISH ECONOMIC POLICY REVIEW 5 (1998) 39-80
Income distribution in Sweden:
what is the achievement of the welfare state?
Anders ~ j ~ r k l u n d *
Summary
By tradition, most studies of income distribution have focused on
cross-sectional inequality of annual disposable income. The paper
starts by summarizing conclusions that can be drawn from such
studies. Even in the midst of the deep recession in 1992, Sweden had
retained its position among the four or five OECD countries with
the highest degree of equality. The paper continues by reviewing two
rapidly growing fields of literature on long-run equality and on equality of opportunity as measured by intergenerational income correlations. The conclusions about successful outcomes regarding equity
are not changed when these perhaps more basic dimensions of
equality are used. i~
* Professor of Economics at the Swedish Institute o f Social Research at Stockholm Univerigl and
a member fthe Swedish Economic C o ~ n dHis
. main jelds of research are labor economics,
income distn'bzltion, and economics ofthe,farnib.
39
SWEDISH ECONOMIC POLICY REVIEW 5 (1998) 39-80
Income distribution in Sweden:
what is the achievement of
the mlfare state?
Anders Bj orklund'
There is no doubt that the major goal of the traditional Swedish welfare state has been to equalize economic outcomes among the residents of the country. Income taxes have been high and progressive.
A variety of transfer systems have been introduced to help families
with low incomes. This is not to say that each public transfer system
has been designed to maximize its immediate equalizing impact on
income distribution. A typical feature of Swedish social insurances
has been to rely on the principle of income replacement. This is basically the design of unemployment insurance, sickness benefits, maternity leave, and pension systems. Further, support to families with
children has been quite general in nature rather than means tested.
Child allowances, for example, are universal and paid to all parents
irrespective of their income and wealth.
It is often argued that the principles of income replacement and
universality are parts of an ambitious overall strategy to achieve a
high level of equality. Public social insurance, based on the incomereplacement principle, might crowd out private insurance schemes with
less egalitarian outcomes as a consequence. And universality of the
systems might increase their political support among the middle class
and even among those in the upper part of the income distribution.'
Yet another characteristic of the Swedish approach to welfarestate policy is the strong emphasis on universal access to public services, such as education, health-care services, and care of the elderly. A
popular expression among proponents of this approach is that the
I am gratful to lohan FtitzelI, Markus Jantti, Mats Persson, Birgtta Swedenbotg and an
anonymow rej%reefor usfuL di.rcussion and to Annita Nasstrom for making the jgures. The
pqer summaizes research financed @ the Nordic Councilfor Economic Research, the Swedish
Coundfor Sorial Research,the SweSsh Council far Research in the Humanities and S o d
Sciences, and The Bank of Sweden Tercentena~Fogndation.
*
I<orpi and Palme (1998), for example, have forcefully advanced these arguments.
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
size of the wallet should not affect availability of such fundamental
services. For example, regarding education, this view can be interpreted as preferences for equality of opportunity: the sixe oftbeparents'
walIets should not afect the educational choices @children.
The ambition to equalize pre-tax and transfer outcomes in the labor market also illustrates the emphasis on equality in Sweden.
Swedish unions have been strong and have tried to reduce wage differentials. In practice, the solidar$ wagepoliq has been a policy to reduce wage differentials. A motivation for public labor market policies
has been to support this wage policy of the unions.' In all, the Swedish people have revealed preferences in favor of equality in several
dimensions.
In the 1990s, the Swedish welfare state has been challenged in
several ways, the most important one being the big budget deficits.
With the high overall tax rates, it is not surprising that expenditure
cuts have been the main candidates for measures to reduce the deficit. Another threat to the equality achieved in Sweden has been the
high unemployment rate, which seems to have become permanent.
European integration and more intense international competition are
also regarded as (future) threats to the equal labor market outcomes
in Sweden.
In this economic environment, it is inevitable that participants in
the public discussion argue that the very ambitious goals of equality
must be reexamined. It is also argued that there are some more basic
dimensions of equality that are more important to protect and some
less important ones. A common claim is that public services are more
important than public transfers to achieve the basic goal of equal opportunity. Another common claim has been that the most important
goal is h)time equality, whereas the ambitions to equalize individuals'
incomes over the life cycle could be reduced.
No matter what is true about these future challenges of the welfare state, it is important to identify the:
Goals of equality that the Swedish welfare state achieved up to
around 1990
Extent of the setbacks that have occurred during the 1990s
The purpose of this paper is to address these two issues.
See Edin and Holmlund (1995) for an overview of solidarity wage policy and
labor market policy.
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
In looking for empirical evidence to evaluate the extent to which
the Swedish experiment has been successful in equalizing economic
outcomes, the most detailed information pertains to income measured over a single year. Like the statistical offices in many other
countries, Statistics Sweden runs an Income Distribution Szlwey, which
focuses on the distribution of annual income. Thanks to the work of
Statistics Sweden, the quality of these data has improved over time.
The data are also quite rich in the sense that they provide information about several income concepts. At the international level, the
Luxembozl~ Income S t u 4 (L,IS) has collected national incomedistribution surveys from many countries into one multi-national research d a t a b a ~ e .Statistics
~
Sweden's Income Distribution Suwy is the
Swedish data contribution to LIS. The LIS organization has also
made large efforts to improve comparability among countries. For
these reasons, there is now growing literature on cross-country comparisons of inequality of annual income. In these comparisons, Sweden has, in general, come out as a country with one of the most equal
distributions of annual disposable income.
But, the focus on income received in a single year has been criticized as a too narrow measure to determine how successful a country
has been in achieving economic equality and justice. The most obvious limitation is probably that a year is a short period. Annual income can be thought of as containing at least three components:
1. A lifetime or pe~manelztcomponent.
2. A component that represents the phase of the life cycle that the
individual is in at the m o m e n t a student, newly married (or cohabiting) without children, raising children, living with a working
spouse without children living home, and being retired.
3. A transitory component, which captures the fact that the individual in all phases of life is exposed to shocks such as sickness, unemployment, and simply bad or good luck. Transitory income
variation can also appear by voluntary choice if, for example, a
person works very hard for a couple of years to save money for a
later period of leave of absence to travel around the world.
By using annual data, the researcher attaches equal weights to all income differentials irrespective of the nature of the income differentials. Another obvious limitation of data on inequality of economic
3
See Atkinson, Rainwater and Smeeding (1995) for detailed information.
INCOME DISTRIBUTION IN SWEDEN, Anders BjorMund
outcomes, annual or lifetime, is that they in themselves cannot be
interpreted regarding inequality of opportunity.
I continue the paper by summarizing evidence from traditional
annual data on disposable income. Section 1 discusses data and
measurement issues. In Section 2, I show the evolution of Swedish
income distribution using annual data from 1975 to 1995. Section 3
presents the Swedish record from a cross-national perspective,
mainly using results from studies based on LIS. Section 4 examines
whether the conclusion that Sweden has a successful outcome regarding equality is changed if broader measures of income are used
and if equality of opportunity is used as the criterion of success. Most
of Section 4 is devoted to studies that use data from time units longer
than one year. Section 5 summarizes evidence from rapidly expanding literature on intergenerational earnings mobility as measured by
correlations between earnings of fathers and sons. Section 6 concludes the paper.
1. Issues of measurement and methodology
1.1. Measurement
A common starting point for income measurement over a certain
period is the Haig-Simons definition of income. According to this
definition, income is defined as the maximum consumption a person
can afford in the period without ending the period with lower net
wealth than at the start, that is, without reducing future consumption
possibilities. This also means that income is actual consumption in
the period plus the change (positive or negative) in net wealth during
the period. Because income is defined in terms of potential consumption, it is obvious that both cash and non-cash components of
income ideally should be included.
Even though the Haig-Simons definition offers some general
guidelines for measuring income, several practical problems appear in
empirical work. I follow what today is the most frequent approach to
measuring disposable income in applied income distribution research.
Income is measured with the household as the unit @income, which
means that post-tax and transfer income of all persons in the household are included. This measure of total household income is divided
by the equivalent number afadults in thefamib, using an equivalence scale.
For example, the income per person, obtained in this way, is distrib-
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
uted equally among all members of the household, an assumption
about equal sham'ngwithin the household. This assumption implies that
inequality between men and women is ignored in this kind of analysis. Gender inequality obviously requires another analysis. In the final
step, inequality is measured among all individuals in the total sample;
so the individual is the unit ofanabsis. A nice property of this procedure is that a measure of the economic standard of children is also
obtained, so inequality among children can also be measured.
The choice of equivalence scale in this procedure is crucial. The
applied scales generally have two properties. First, children are generally considered ajnancial burden for the adults in the household, even
though this assumption is not necessary to make. Second, most scales
imply some economies of scale within the household, so the marginal
cost of additional adults (and additional children) falls with the number of adults (or children) in the household.
When these general principles are applied on data from Statistics
Sweden's Income Distm'bution Sumey, some weaknesses of the approach
should be kept in mind. First, the survey applies a rather narrow
definition of the household. Only two spouses are considered adults
in a particular household, even if more adults live in the same household. Additional household members are counted as separate households. For example, all adult children, age 18 and older, who live with
their parents and grandparents, who in some cases live in the same
household, are counted as separate household^.^ Second, capital income is quite poorly measured. Capital gains are included only when
they are realized and to the extent that the gains are considered income in tax assessments. Further, if no special correction is made, all
capital income is measured in nominal terms. Third, there are no attempts to include consumption of various free public services in income, even though some such services are close substitutes to cash
income. Fourth, almost all income data stem from employers' reports
to tax authorities. Obviously, such a measure of reported income
does not capture income from the hidden sector of the economy."
Households with members from three generations of a family are rare. According
to the LevelofLiving Suwy of 1991, less than one-half of one percent of residents in
Sweden, ages 18-75, lived in such households.
5 The issue about the discrepancy between reported and real income due to tax
evasion seems to be neglected in the empirical literature. Persson and WissCn
(1984) analyze this problem by means of a theoretical model and can characterize
the conditions under which the h - o distributions differ from each other in various
ways.
4
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
Finally, the Income Distribution Sum9 is a sample of about 10,000
households, so standard errors of inequality measures might be of
non-trivial size, especially for subgroups of the population.
1.2. Methods
To find out how the welfare state by means of taxes and transfers has
affected the income distribution, I need a counterfactual that describes what income distribution would have been without these
policies. In this paper, I follow a long tradition that simply compares
the actual distribution of disposable income with the distribution of
pre-tax and transfer income. Although this procedure might yield a
reasonable approximation to the relevant counterfactual, it is important to stress two limitations of it. First, one must ignore economic
behavior, such as labor supply and savings, which taxes and transfers
might affect. The same holds for prices and wages that are assumed
to be unaffected by welfare state policies. A very elaborate behavioral
model of the economy would be needed to take such effects into account.6 In my opinion, there is no such model available in the literature. Second, it might be unrealistic to assume that the counterfactual
to the welfare state does not contain the protection that some of the
public social insurance systems offer. In particular, it is reasonable to
assume that private pension systems would grow in importance without a public one. Also, the private market, without public support,
would probably offer some private sickness-benefit systems and even
unemployment benefit schemes.
Fully aware of these problems and urithout access to the required
behavioral models, I confine myself to analyzing the impact of taxes
and non-taxable transfers. The most important of the latter are child
and housing allowances, social assistance benefits, and advance
maintenance for single parents. Even though both income taxes and
such transfers might affect economic behavior, a private market
would not offer the transfers in the absence of a public welfare state.7
Because disposable income is measured with the family as the unit of income and
the family size is adjusted for by equivalence scales (see Section 1.1),not only economic models of labor supply behavior and price determination are needed, but
also demographic models of mating, fertility, and divorce.
Maybe there would be more, and more active, charity organizations without such
public transfers.
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
2. The evolution of inequality s f annual income
2.1. 1975-199kthe Income Distribution Survey
The Income Distribution Suwy offers time-series information on income distribution from 1975-1995 (with comparable data missing for
1976-77 and 1979).' Figures l a and 1b show the evolution of inequality over this period for three groups, namely: all persons (including
children), prime-age adults, ages 30-54, and children, ages 0-17. From
a general viewpoint, it is appealing to study income inequality among
all individuals in society. But because of the weakness of the household concept in the Income Distr;;bution Sut7y and the fact that it is
commonly accepted that high school9 and college students can live
relatively well on low incomes, I also use an age group that excludes
most students. For example, in looking at ages 30-54, I eliminate
most students and focus on a group that the labor market strongly
affects. Children are interesting in their own right. The economic
standard of children is not, in any respect, voluntarily chosen by the
child, and equality of economic standard among children might reflect equality of opportunity. The equivalence scale that I use is
commonly used in Swedish studies and is based on official guidelines
about social assistance for families of different sizes. In this scale, the
first adult gets the weight 1.0; the second adult, 0.65; a child age 0-3,
0.35; a child age 4-10, 0.43; and a child age 11-17, 0.52.
Figure l a shows two commonly used measures of overall inequality, the Gini coefficient and the 90110 percentile ratio. The general
conclusions drawn are not sensitive to this choice of measures of
inequality and equivalence scale.'' Figure 1b contains the 90150 and
10/50 percentile ratios and consequently provides information about
the inequality within the lower part (10150) and within the upper part
(90150) of the distribution. Note that a high value of the 10/50 percentile ratio implies a high degree of equality, whereas a high value of
the 90/50 percentile ratio implies the opposite.
Gustafsson and Palmer (1997) and Ministry- of Finance (1996, 1997) offer alternative recent presentations of the evolution of Swedish income distribution using
the same basic data set. Contrary to Statistics Sweden, Gustafsson and Palmer add
2.5 percent of the net asset value of an owner-occupied home to income.
The normal graduation age from high school is 19 years in Sweden.
lo Figures that use alternative measures of inequality and an alternative equivalence
scale are available from the author upon request.
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
INCOME DISTRIBUTION IN SWEDEN, Anders Rjorklund
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
INCOME DISTRIBUTION IN SWEDEN, Anders Blorklund
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
It is natural to start by looking at the 1975-1990 period, and then to
turn to the turbulent 1990s. Stability is the most striking pattern in
the evolution of inequality in this period. For all three groups, the
Gini coefficient fluctuates within two percentage points. For the entire population, the range in the period is 0.20-0.22. The 90/10 percentile ratio fluctuated between 2.40-2.60 for the entire population.
The fluctuations for the more homogenous group (ages 30-54) are
slightly lower. But for children, there is an equalizing trend until 1987,
after which it is reversed. These fluctuations are modest and motivate
stability as the label for this period, considering standard errors of
Gini coefficients on Swedish data, which are around 0.004 and considering the magnitude of cross-country differences in Gini coefficients and 90/10 ratios, which are reported below." From the methodologcal viewpoint, it is interesting to note that the 1986 spike in
the Gini coefficient does not show up in the 90/10 ratio. The reason
is most likely that there are a few observations with very high income
in the very top of the 1986 distribution.
There is no obvious peak in equality in the period. For the entire
group, 1981 is the peak according to both measures of inequality,
whereas the peak for prime age and for children came later in the
1980s.
Figure 1b can tell whether the apparent stability of income inequality over the period reflects stability in both the upper and the
lower part of the distribution, or if there were counteracting forces
involved. The answer seems to be that inequality has been quite stable in both parts of the distribution.
The 1990s have been turbulent in several ways from the incomedistribution viewpoint. The recession with very high unemployment
started in 1991, and since then unemployment rates have been very
high by historical Swedish standards. Partly as a consequence of rising
unemployment rates and partly due to educational reforms, the
school enrollment rates among ages 18-25 increased quicMy over the
1990s. A large tax reform, with lower marginal tax rates, broader
bases of taxation and higher transfers to families with children, was
implemented in 1991. In the next years, budget cuts with lower replacement rates in most social insurance schemes and reduced uni11 It is not easy to say what is much and what is little in characterizing changes and
differences in inequality. The reader can use the property that the Gini coefficient
equals half the expected percentage difference between two randomly drawn individuals in the population that is studied.
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
versa1 child allowances were made effective. Further, the rate of inflation fell markedly with lower discrepancy between real and nominal capital income as a consequence.
The economy changed in several ways and so did the data that are
used to analyze income distribution. First, the broadening of the tax
base in 1991 changed the income concepts in several ways. Second,
due to changes in the incentives to sell stock, realized capital gains
were unusually high in 1991 and 1994. And finally, as previously
mentioned, inflation fell markedly with consequences for the content
of capital income.
To avoid the first of these problems, I use revisions of income
data for 1989 and 1990, made by Statistics Sweden, to achieve comparability over time. As shown in Figure la, the Gini of these revised
income data is about one percentage point higher for 1989 and 1990,
so the income definition seems to have raised the Gini coefficient by
this magnitude. But some caution is called for, because it was not an
easy task to determine how much the broadening of the tax bases
affected measures of income.'' Regarding the second problem, I see
no other way to solve it than to treat the observations for 1991 and
1994 as outliers due to temporarily high capital gains in these years.13
But this could underestimate inequality in the surrounding years if
unequally distributed capital gains were reallocated over time to the
years with low tax rates. I have made no efforts to solve the third
problem either, but refer to Section 4.4, where I report some results
based on alternative definitions of capital income.
Considering these cautionary notes, we can now look at the evolution of income inequality in the 1990s. For all and for ages 30-54,
there is definitely an increase in income inequality from 1990 to 1995
according to both measures of overall income inequality. The increase is also slightly larger for all persons than for prime age. This
could reflect that a rising number of young students are included in
the first group but not in the second. But the increase from 1990 to
1995 is only modest (if we can trust the revised numbers for 1990);
the Gini increased by at most one percentage point and the 90/10
ratio from 2.58 to 2.78 for all persons (and slightly less for ages 3012 See Bjorklund, Palme, and Svensson (1995a) for more infornabon and a discussion.
' 3 1994 was the most extreme of these years. Realized capital gains increased from
17 billion Swedish crowns in 1993 to 59 billion in 1994 and fell back to 20 billion
in 1995 (Statistics Sweden 1997).
INCOME DISTRIBUTION IN SWEDEN, Anders Bjiirklund
54). Given the dramatic increase in unemployment and cuts in replacement rates in most social insurance systems, it is surprising that
inequality did not rise even more.14 For children, the stability in inequality in the 1990s is even more remarkable. The Gini coefficient
has not changed from 1990 to 1995 and the 90110 ratio reveals only a
minor increase in inequality.
Figure l b shows that the rise in overall inequality as measured by
the 90110 ratio can be attributed to changes in both parts of the distribution. The next question is how taxes and transfers have contributed to the evolution of income distribution.
For the three demographic groups, Figure 2 shows the Gini coefficient of:
1. Market income (income before taxes and tax-free transfers)
2. Market income minus taxes
3. Market income minus taxes plus transfers
The difference between item 1 and item 2 illustrates the equalizing
impact from taxes. The difference between item 2 and item 3 illustrates the equalizing transfers impact.15 As expected, the figure reveals a marked equalizing impact from taxes and tax-free transfers,
and the mechanicaI, redistributive, transfers impact is somewhat higher
than the taxes impact.
A noteworthy change occurred in the 1990s. The overall taxes and
transfers impact has increased for all groups, and there is a clearly
rising transfers impact, in particular for children.
For 1991 and 1992, the increase in child allowances, which was
part of the tax reform, is the most likely explanation for this development (see Bjorklund, Palme and Svensson, 1995b).
In 1995, child allowances were reduced again, so probably rising
social assistance payments help explain that the equalizing transfers
impact continued to increase in 1995.
l4 The Finnish experience in the early 1990s better illustrates that drastically rising
unemployment need not lead to immediate large increases in income inequality in
countries such as the Nordic ones; see Aaberge et al. (1997).
l5 It is hard to determine if taxes should be deducted before transfers are added, or
if the opposite order should be used, but this choice does not qualitatively affect
the conclusions. Figures are available from the author on request.
Figure 2. The equalizing impact of taxes and transfers. Gini coefficient of
market income before taxes and tax-free transfers (MI), market income
minus taxes (MI taxes), and disposable income (Dl).
-
All
Sowce: Statistics Sweden's Income Distribztion S~rvey
INCOME DISTRIBUTION IN SWEDEN,Anders Bjorklund
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
2.2. The longer run
Because the Swedish welfare state expanded during the 1960s and
early 1970s,16 we would like to know how income distribution
evolved during the period when the welfare state expanded. T o infer
that the equal distribution of income observed in the early 1980s is
due to the build-up of the welfare state, we should observe falling
income inecpality during the 1960s and early 1970s.
Unfortunately, the information on income distribution before
1975 is meager. But there are two useful observations on inequality of
disposable income for 1967 and 1980 from the Swedish Level ofLiving
Szlwys. Several authors have used this data set.17 There are two striking patterns in the results. First, overall inequality declined markedly
from 1967 to 1980 for several choices of equivalence scales and
measures of inequality. The Gini coefficient, for example, declined by
almost 10 percentage points, a very large number in light of the
country differences reported in Section 3 and changes from 1975 to
1995 reported in Section 2. Second, the redistributive effect of taxes
and transfers increased markedly from 1967 to 1980.18These findings
strongly support the view that the expansion of the public sector
contributed to a more equal distribution of disposable income.
Even though the Level ofLzving Suweys in all respects is a highquality data set, these findings must be interpreted with some care.
The main reason for this is that there is only one observation from
the 1960s, and as we previously explained, Swedish data series on disposable income tend to be somewhat erratic. If 1967 was an unusual
year, like 1994, the numbers might be misleading. But the large decline in inequality of hourly earnings from 1968 to 1981 is consistent
with the evolution of inequality of disposable income (see Edin and
Holmlund, 1995).
16 Total public spenhng as a percentage of GDP expanded from 30% in 1960 to
about 50% in 1975 and further to almost 70% in the early 1980s. See, for example,
Lindbeck (1997) for a recent review of the rise of the Swedish welfare state.
17 For example, Fritzell (1991), Gustafsson and Uusitalo (1990), Jansson (1990),
and Aberg (1989).
This finding must be interpreted with some care. All types of transfers were
included in the analysis. Because the pension age was reduced during the period, it
is not surprising that mechanical comparisons of pre- and post-transfer income
suggest rising redistributive effects. A complete distributional analysis of pension
systems requires a comparison between present values of contributions and payments (see Stihlberg, 1990).
INCOME DISTRIBUTION IK SSWEDE,N, Anders Bjorklund
3. Cross-country compaksons-evidenee from LIS
T o compare income inequality in Sweden with other developed
countries, I rely on results from studies based on the L,.uxemboug Income St24 (LIS). The definition of disposable income in LIS differs
from the one used by Statistics Sweden in some waJ.s." The two
most important ones are that capital gains and student loans are
taken out by LIS and not considered part of income. The reason for
doing so is that this information is missing for some other countries,
so cross-country comparability is in this way increased.
It is relevant to ask if one can expect any systematic bias in the estimates of inequality in Sweden compared to other countries in LIS.
One argument in favor of an upward bias of inequality in Sweden is
the fact that persons from the age of 18, who remain with their parents, are counted as separate households. Because they often are students with no or very low incomes, they are also counted as persons
(and households) with low incomes, even though they most likely
benefit from the economic resources of their parents. For the same
reason, too high incomes are attributed to parents with children of
this age who have not yet moved to their own housing. So some care
is called for when interpreting Swedish inequality estimates based on
samples in which youth between ages 18 and (say) 25 are included.
Another problem pertains to the rise in inequality in Sweden in
1991, when, as previously discussed, Swedish income concepts became broader and measured inequality rose as a consequence. My
guess would be that this change implied that some unequally distributed components of income are measured more accurately in Sweden
than in other countries. So in all, inequality estimates for Sweden are
probably slightly overestimated, and the overestimation is probably
higher after 1991. Further, the overestimation becomes higher, the
higher the weight for young people in the sample investigated.
With these cautionary notes in mind, we now turn to some central
results from studies based on 1,IS. I first focus on international comparisons of overall inequality and then turn to studies of specific vulne~ablegroups. Table l a shows Gini coefficients and 90/10 percentile
ratios for several countries in the LIS data set during the period from
around 1980 to the early 1990s. Table 1b is similarly organized and
shows 90/50 and 10/50 percentile ratios. The square root of the
l-ee
Atkinson et al. (1995) for Inore detailed lnfonnation on coinparabill?among countries.
INCOI\IIE DISTRIBUTION IN SWEDEN, Anders Bjorklund
number of persons in the household is used as an equivalence scale,
and inequality is measured among individuals with the assumption of
equal sharing within the household. No age restrictions are used, and
children are also included.
_As Table l a shows, Sweden had the most equal distribution of
disposable income in the early 1980s, when 10 countries are included
in the comparison. In the late 1980s and early 1990s, when more
countries are included, Finland tops the list; Sweden is second. But
the differences between the five countries with most equal distributions are quite small. In the early 1990s, these are the four Nordic
countries and Belgium. In all periods, the U.S. had the most unequal
distribution, but in the early 1990s the UI< is not far behind.
Table l b helps us determine to what extent the lower or the upper
part of the distribution explains the outcomes for the various countries. The table gives the general impression that countries with low
overall inequality have low dispersion in both parts of the distribution. And the high-inequality countries (UI< and U.S.) have high dispersion in both parts of the distribution.
Many international comparisons of income distributions have
particularly focused on specific vulxerable groups that historically have
had high poverty rates and are in greater need of welfare-state arrangements than others. A general result from such studies has been
that Sweden has done well in equalizing income and eliminating poverty among such groups. I<orpi and Palme (1998, forthcoming) report that Sw-eden had the lowest overall inequality (measured by the
Gini coefficient) and the second lowest poverty rate for elderly (ages
65 and older) among 11 OECD countries around 1985. Coder et al.
(1989) and Fritzell (1992) get similar results for the period around
1980.
Comparative studies, which have focused on the income standard
of children, also reveal favorable outcomes for Sweden regarding
equality. Studies by Coder et al. (1989), Fritzell (1992), Jantti and
Danziger (1994), and Rainwater and Smeeding (1995) have shown
four striking results for children in Sweden: Overall inequality is very
low by international standards. Poverty rates are low. The absolute
income level among the poorest Swedish children is high. These patterns hold for children living in both single and two-parent families.
INCOME DISTRIBUTION IN SWEDEN, Anders Bjijrklund
Table l a . Inequality of disposable income in OECD countries in the
early 1980s, late 1980s, and early 1990s. Gini coefficient and
P90lP10. Ranks by equality.
Israel
.305 1992
Germany
-
U.S.
4.93
1979
10
3.00
1984
6
5.94
1986
14
9
-
5.78
1991
13
lence scale. 2) Standard errors of the ('Jini coefficients are not available. For Sweden they tend to be around 0.004 for the Gini coefficient.
Socl~~ces:
L%tkinsonet al. (1995) and ('Jottschalk and Smeedirlg (1997).
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
Table 1b. inequality of disposable income in various countries in
the early 1980s, late 1980s, and early 1990s. P90lP50 and P I 0lP50.
Ranks by equality.
Notes: See notes for Table l a .
INCOME DISTRIBUTION IN SWEDEN, Anders Bjiirklund
The evidence in Sections 2 and 3 can be summarized in six conclusions:
1. Together with the other Nordic countries and Belgium, Sweden
had the highest equality of annual disposable income among 1014 OECD countries from around 1980 to the early 1990s. This
result is quite robust and holds for most measures of inequality,
and for most equivalence scales.
2. In an international perspective, equality was high and poverty
rates low among vulnerable groups, such as the elderly, children,
and single-parent families in Sweden during the same period.
3. The time-series information provided by Statistics Sweden's Income Distm'bution Sumy, which covers the 1975-1995 period, shows
that equality peaked in the early 1980s, but the rise in inequality
since then has been modest.
4. The turbulent period of the 1990s is somewhat hard to characterize due to changes in income definitions, high realized capital
gains in 1991 and 1994, and divergent patterns for various age
groups. Among adults (ages 30-54) and children (ages 0-17), inequality remained remarkably stable despite the drastic increase in
unemployment.
5. Most likely, a marked equalization of disposable income occurred
from the mid 1960s to around 1980 when welfare-state policies
were expanded and the government share of GDP increased
rapidly.
6. Income taxes and tax-free transfers have had strong equalizing
impacts on the distribution of disposable income.
4. How robust is the evidence?
How robust are these conclusions based on annual data? Are they
only artifacts from using a single year as the time unit or from the
specific income definitions that are employed? I start the discussion
of these questions by reviewing results from studies that use data for
longer time periods.
4.1. The time period
With perfect capital markets and perfect foresight, the individual can
consume out of her lifetime income, independent of the time path of
income. Lifetime income will be the sole determinant of &time atility.
Rut without access to capital markets, fluctuations in the income
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
stream are also likely to affect lifetime utility. Assuming concave inter-temporal utility hnctions, it follows that the individual prefers an
even stream of consumption (when family size is adjusted for). So
fluctuations of income (adjusted for family size) will reduce lifetime
utility. In general, it seems reasonable to argue that lifetime utility is a
long-run or permanent-income
positive function of lifetim-r
and a negative function of income fluctuations. The better the individual can smooth her consumption path by means of the capital
market, the less weight should be attached to fluctuations.
But to start, let us assume that we are willing to attach all weight
to long-run income and ignore variability of income paths. Would the
previously stated conclusions about successfL1 outcomes of welfarestate policy, drawn from annual data, be changed? Apm'om', there are
quite strong reasons to believe that distributions of lifetime income
are markedly different from distributions of annual income. The early
study by Blomquist (1981), in which he simulated distributions of
lifetime income by using wage and hours equations estimated on longitudinal data for 1968 and 1974, suggests that inequality of lifetime
income is 40-50% lower than annual income. Bjorklund (1993) used
actual longitudinal data and compared inequality of market income of
adult men over the 1951-89 period and found that inequality of income over this long period was 35-40'70 lower than cross-sectional
inequality of annual income in the same period.
Another indication of high mobility of income, and thus a discrepancy between inequality of annual and more long-term inequality,
comes from studies of transition matrixes of income in two time periods. Fritzell (1990) found that almost half of those in the lowest
quartile of the distribution of family-sized, adjusted, disposable income in 1997 had moved to a higher quintile seven years later. In a
recent study, which has received attention in Swedish media,
Uddhammar (1997) reports patterns of mobility in disposable income
from 1985 to 1991. He focuses upon those 5.6% of the population
with disposable income below 50% of the median income in 1985.
'This group seems to be particularly mobile because only 190/o had
very low incomes also six years later.20
More information is needed to reconcile the Fritzell and Uddhammar tindings.
But it seems as though those at the very bottom of the income distribution are
particularly mobile. The conditional probability of those at the very bottom of the
income distribution in 1985 to be in the top of the distribution in 1991 is higher
than for those next to the bottom.
20
INCOME DISTRIBUTION IN SWEDEN, Anders Bjijrklund
Thanks to better access to longtudinal data sets in which individuals have been followed over longer periods than one year, there
is now a rapidly expanding literature that focuses on income distribution in much longer time perspectives than a single year. A few recent studies have employed longitudinal data sets from several countries, have defined as similar income concepts and samples as possible, and have compared inequality of annual and long-run income
among countries. Because of the properties of standard inequality
measures, inequality of annual and long-run income will be the same
if relative incomes of all individuals remain constant in each period.21
The more changes of relative incomes that there are, the lower inequality of long-run income will be. In principle, changes of relative
income can be so frequent and of such magnitudes that long-run income is completely evenly distributed. Changes in relative incomes
over time, which reduce long-run inequality, can be called income
mobility.
Table 2 summarizes results from three such studies that have focused upon inequality of labor earnings of individuals, that is, the
nature of pre-tax inequality, which the labor market generates.22
As expected, the results in all three studies confirm that the U.S. is
the country with highest annual inequality. But what is striking is that
prolongng the time period does not bring the U.S. much closer to
European levels of inequality. Somewhat surprisingly, Germany has a
higher degree of mobility than the U.S. in the two studies where both
countries are included. The results for Sweden and the U.S. in the
study by Aaberge et al. (1996) can illustrate that the differences in
mobility between the countries are quite small: the Swedish Gini for
annual income is 65.6% of the corresponding U.S. Gini, whereas the
Swedish Gini for long-run income is 68.4% of the U.S. Gini. Results
in O E C D (1996), based on the transition-matrix approach, support
the view that the similarity of earnings mobilit-y patterns across
countries are more striking than the differences.
The Gini coefficient has a slightly different property in this respect; the Gini of
income in single periods equals the Gini of average income over longer periods if
the units (here the individuals) in the sample do not change rank. See Shorrocks
(1978).
22 The exact definitions of labor eamings and other choices involved in studies of
eamings and income distribution differ among the studies but are as similar as
possible for each country in each study. More detailed information about measurement and more results (including various sensitivity analyses) car1 be found in
the original papers.
21
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
Table 2. Annual inequality, long-run inequality, and percentage reduction in inequality from extending the time period.
Pre-tax earnings of individuals. Results from three studies.
Study
Country
Inequality Long-run Average
measure
time
annual
Long-run
ineaualitv
Percentage
reduction.
Note: The original studies contain more detailed information about data sources,
income concepts, *andsample criteria.
a A. = Aaberge et al.
B. & P. = Burkhauser and Poupore
Table 3 is organized in the same way as Table 2 and shows similar
results for disposable income adjusted for family size.23The overall
conclusions are about the same as for earnings. The popular view
that inequality in the U.S. is largely attributable to a high degree of
income mobility gets no support at all.24
So an overall result from these studies is that mobility of earnlngs
and incomes during periods up to 11 years are strililngly similar
across countries. So the ranking of countries regarding inequality is
not very much affected by the length of the time period that is used.
Conclusion 1-that
Sweden internationally ranks high regarding
23 Due to comparability problems, the measures of disposable income are less
complete than the ones used in Sections 2 and 3.
24 Fritzell (1990), using the transition matrix approach, 1973 to 1980 for Sweden
and 1971 to 1978 for the U.S., also finds remarkably similar mobility of disposable
income in the two countries.
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
equality-does not seem to be an artifact of using annual data. The
results also imply that the relatively high inequality that is observed in
the U.S. is attributable to permanent inequality, not to higher mobility of earnings and income among individuals and families.
Nonetheless, some resen-ations are in order. The results might be
due to the fact that it has not been possible to use data from complete life cycles but only time periods up to 11 years long.25And even
though the authors of these studies have made large efforts to
achieve comparability among countries, it is more difficult to achieve
such comparability for longer time periods than for a single year.
Some information in recent studies also helps us determine
whether conclusion 6 is an artifact of using annual data, that is,
whether the equalizing impact of taxes and transfers vanishes when
the time period is extended. The studies by Aaberge et al. (1996) and
Burkhauser and Poupore (1997) also contain information about inequality of market income (pre-tax and transfer income) and disposable income (post-tax and transfer income) for both single years and
for the longer time periods in Table 3. For all five countries included
in the two studies, the percentage reduction of inequality was practically the same for a single year and for the long time periods applied
in the studies. So at least the combined effects of taxes and transfers
do not seem to disappear when the time period is prolonged.
But it is a strong assumption to attach all weight to long-run income and completely ignore individual income variability that contributes to cross-sectional inequality. To determine how important
the latter is, information about the nature of income variability is useful. If disposable incomes fluctuate for the wealthy, for example, because of capital gains, then variability hardly represents welfare losses
25 Even though such a resenration is motivated, some results suggest that the impact of the length of the period on the percentage reduction of inequality declines
markedly after 5-10 years, (Gustafsson, 1994 and Zandoaliili and Gustafsson,
1997). This finding is not inconsistent with the previously discussed results-that
inequality of lifetime income is around 40% lower than cross-sectional inequality
of the population. The reason is that cross-sectional inequality of annual income
includes the impact of inequalities anlong cohorts, whereas the longitudinal studies
of income mobility are done for a limited number of cohorts.
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
Table 3. Annual inequality, long-run inequality, and percentage reduction in inequality from extending the time period.
Disposable income of families. Results from two studies.
Study
Country
B. & P.
(1997)
German
U.S.
Inequality
measure
Gini
Long-run Average
time
annual
period
inequality
1986-90
.341
bong-run
inequality
Percentage
reduction
mobility
.321
6.0
Note: See notes for Table 2.
a A. = Aaberge et al.
b B. & P. = Burkhauser and Poupore
for these people and does not cause severe inequities. An uneven income path due to college studies during young age, followed by high
and stable incomes as an adult, can hardly be considered problematic
either. If instead, incomes fluctuate strongly for those with low longrun income, the variability might represent a welfare loss. Inequality
of long-run income would in this case underestimate inequality of
lifetime utility.
A recent study by Bjorklund and Palme (1997) on Swedish data
provides some information about the relationship between long-run
income over 18 years and inter-temporal income variability for single
individuals during the same period. For this long time period, they
compute inequality of long-run income among individuals and intertemporal income variability for each individual in the sample using
the same measure of inequality. With this information, it is possible
to characterize income variability.
Table 4 shows the magnitude of income variability by quartiles in
the distribution of long-run income. Results are reported for Theil-0
and Theil-1 measures of inequality; the former measure is more sen-
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
sitive to low incomes. Also note that the Theil measures, like most
measures of income inequality, capture relative inequalities, so in this
case, the individual's income variability is measured relative to his
long-run level of income. Results are also reported for two samples:
one that includes youth plus young adults and another with only
prime-age adults.
One conclusion from the results in the table is that variability of
household market income (pre-tax and transfer income adjusted for
family size) is highest for those with low long-run disposable income.
This pattern holds for both samples. In the adult sample, income
variability is higher in the top quartile than in the second and third.
So the general pattern is that those who suffer most from low longrun income also suffer the most from income variability. Most likely,
those with low long-run income have less access to a perfect capital
market than those with high long-run income. It follows that income
variability cannot be neglected in a study of income inequalities, even
if lifetime utility is the overall norm for equity.
A second result from the table is that the equalizing impact on income variability of taxes and transfers is highest for those with low
long-run income.26So the welfare state seems to do a good job in
stabilizing income paths for those who need it most.
The study also examined the extent to which taxes and single
transfer schemes affect the tsvo dimensions of inequality, the longrun component and the component that shows variability. It is possible that some transfers, at least for some measures of inequality and
for some equivalence scales, only affect income variability but not
inequality of long-run income or vice versa. Indeed, it is even possible that for some transfer schemes, a conflict is involved so that, for
example, income variability is reduced but inequality of long-run income is increased. The results for the separate impact of taxes, universal child allowances and means-tested housing allowances for the
two samples, the two Theil measures of inequality, and for two
equivalence scales suggest that no such conflict is involved. In all
cases, both inequality of long-run income and individuals' income
variability were reduced. So these tcvo goals of income inequality
seem to be complementary and reinforce each other rather than to
be in conflict with each other.
This is true both for the absolute decline in variability and the percentage reduction in variability.
26
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
Table 4. Average income variability In each quartile of
long-run disposable income (1974-1991).
Notes: 1) Theil-0 and Theil-1 are used as measures of variability. 2) The square
root equivalence scale is used in the estimations reported in the table, but the results hold for an equivalence scale that attaches zero weight to children. 3) Income variability is measured as variability around the 18-year average income of
the individual. 'The results are qualitatively the same when variability is measured
as deviations from smooth (second-order) trend. 4) Disposable income is household market income taxes + child and housing allowances.
Sozlrce: Bjorklund and Palme (1997).
4.2. Full income
Another criticism of using current annual income to evaluate how
successful the welfare state has been in achieving economic equity is
that the same weight is attached to differences in income that are due
to variation in working hours as to income differences that are due to
differences in inherent working capacity. This is potentially very misleading in cases when those with lower working hours have specialized in household production or simply have stronger preferences for
leisure. There is no simple way to solve this problem, but the literature offers some attempts to extend the traditional approach in this
direction. For example, Jenkins and O'Leary (1996) have estimated
inequality of household income plus household production and
found that the distribution of such an extended income measure deviated in many respects from the distribution of the traditional income measure.
For Sweden, it would be interesting to know if conclusions 1-6 in
Section 3 would be affected if such a broader concept of income,
also called full income, is used. Bjorklund, Palme, and Svensson
(1995b) attempted to control for differences in working hours by
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
evaluating a standard working time by the wage rate of the individual.
The results suggest that the decline in income inequality from 1967 to
1980 was not as large for hours-adjusted annual income as for actual
annual income. But there were no marked differences between the
two income concepts in the equalizing impact of taxes and transfers.
4.3. Public consumption
Following the Haig-Simons income definition, it is natural to include
the value of public services in a complete measure of income. As regards the choice of time period and the treatment of differences in
working hours, one could expect that the distribution of income
would be strongly affected by inclusion of the value of such services.
Fritzell (1994) made an attempt to estimate the value of child care,
education, health services, and care of the elderly and added these
values to current income. Not surprisingly, the structure of income
distribution was markedly changed.
From an international perspective, it would be interesting to know
whether the conclusion of good Swedish achievement regarding
equality would be affected by inclusion of the value of such services.
Smeeding et al. (1993) made an attempt to include the value of noncash subsidies for health, education, and housing for Australia, Canada, Sweden, West Germany, the Netherlands, UI(, and U.S. The
major result was that such non-cash subsidies rather reinforced the
equalizing impact of cash transfers. So this extension of the traditional income concept did not affect the ranhng of countries.
4.4. Capital income
Many economists argue that the measurement of capital income is
the major weakness in modern income distribution statistics. This
might very well be the case, so it is an important task for the statistical offices to get the additional information that is needed to improve
measurement of such income. But the crucial question is if conclusions, such as those previously drawn, would be markedly different if
capital income were to be better measured. A few attempts have
been made to check if results in Swedish income distribution studies
are sensitive to the definition of capital income.
Bj6rklund, Palme and Svensson (1995a, b) investigated whether
the distributional impact of taxes and transfers, according to the
standard before-and-after comparison, is changed if alternative meas-
INCOME DISTRIBUTION IN SWEDEN, Anders Biorklund
ures of capital income are used. They replaced income from capital
measured by Statistics Sweden by:
8
A 3% real return on estimated market value of wealth
The maximum of income from capital measured by Statistics
Sweden and 3% real return
The separate impacts of taxes and transfers were not changed, but
overall inequality in 1991 was reduced making the evolution of inequality in the 1990s less erratic.
The Ministry of Finance (1996) examined whether the time-series
pattern of Swedish income distribution from 1975 to 1994 is much
affected by the definition of capital income. They adjusted capital
income for inflation by counting only a fraction of nominal capital as
real income, the fraction being determined by one minus the ratio of
the nominal interest rate and CPI inflation. E,ven though this is a
crude adjustment, it is interesting to note that the long-term evolution of income distribution is not very much affected. But the evolution over the 1990s becomes less erratic with lower inequality in 1991
and 1994.
N o doubt more attention should be paid to and more resources
be allocated to the measurement of capital income. Even though
conclusions, such as numbers 1-6, would not necessarily be very
much affected if capital income were to be better measured, the confidence in income distribution statistics among economists would be
raised a lot.
5. Equality of opportunity-intergenerational
earnings mobility
There has also been a strong emphasis on equality of opportunity in
Swedish welfare-state policy: l@'s oppon%nitiesshould not be afected b_y the
siqe fpar-ents' waZZets. It is far from obvious how such a goal can be
made operational. One criterion, which can be implemented empirically, is that the outcome in an individual's life is independent of her
family background. Simple correlations of the outcomes of members
of different generations of the same families, such as fathers and
sons, provide this information. High correlations would suggest low
intergenerational mobility and also a low degree of equality of opportunity.
INCOME DISTRIBIJTION IN SWEDEN, Anders Bjorklund
The study of intergenerational mobility of economic and social
status has a long tradition in sociology. One strand of the sociology
literature has examined class mobility, primarily between fathers and
sons.27Another strand of literature has studied how sociaI statzls is
transmitted between generations.28 Because my focus is on income
inequality, I instead review a rapidly expanding economics literature
on intergenerational correlations in earnings and income^.'^ Within
the economics profession, there has been an interest in such correlations for a long time.
The publication of two studies in 1992, by Gary Solon and David
Zimmerman, on U.S. data in the same issue of the American Economic
Review stimulated more such studies for several countries and several
cross-country comparisons. Solon and Zimmerman found that the
correlations between (the log of) fathers' and sons' long-run incomes
were higher than previously believed: around 0.4, or maybe even as
high as 0.5. Conversely, the results imply that this kind of intergenerational mobility was lower than previously believed.
Solon and Zimmerman also clarified some methodological issues
involved in estimating such correlations from the research data sets
that are available today.30
Table 5 shows a subset of recent studies of such correlation studies. I only include estimates of earnings correlations between fathers
and sons. For several reasons, some care is called for in interpreting
the results; there are both data quality and other methodologcal issues involved in these studies and the literature is still very new. But
there is one pattern that cannot be found in the results. The popular
view in American political rhetoric, that the U.S. is a country that offers a high degree of equality of opportunity that compensates for
high cross-sectional inequality, gets no support in these results.
Erikson and Goldthorpe (1992) is a modem classic in this field of literature.
See, for example, Treiman and Ganzeboom (1990) for an informative survey.
29 The following section is based on Bjorklund and Jantti (1997b) in which more
detailed information as well as a comparison with the sociological literature can be
found.
30 Note that the correlation coefficient between fathers' and sons' log income
equals the elasticity of sons' income regarding fathers' income if there is no longrun trend in the variance of male income.
27
28
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
Table 5. Results from studies of father-son earnings
correlations for various countries.
Author@) and
country
Measure of
earnings
Age of sons
Estimate of
father-son earnings
correlations
(1997)
Wiegand
Comments: For Corak and Heisz: 1) The estimates are elasticities. 2) Non-hear
relationships implying higher mobility at the lower end of the distribution were
found. For Bjorklund and Jintti: the P-value of a one-tailed test of equal correlation was only 0.19. For Couch & Dunn: the low age of sons can explain that correlations are low for both countries. For W~egand:the study is a comment on
Couch and Dunn, and the author claims that his results are more reliable. The
results for the U.S. are not new estimates but taken from Solon (1992).
Notes: 1) For more detailed information, see the original studies and Bjorklund and
Jiintti (1997b). 2) For some studies, the standard errors of the correlation coefficients are as high as 0.10. 3) The range of values refers to results from alternative
techntques to estimate the correlations.
" Average.
O n the contrary, the pattern that can be found in the results goes
in the opposite direction. Both cross-country studies (Sweden-U.S.
and Germany-U.S.) suggest higher correlations for the U.S., even
though the differences might not be statistically significant. Among
the studies of single countries, the U.S. and the UIC, both countries
INCOME DISTRIRTJTION IN SWEDEN, Anders Bjorklund
with high cross-sectional inequality, come out as countries with high
correlations and hence low intergenerational mobility.
It would be premature to argue that cross-sectional equality on
one hand, and equality of opportunity on the other, are complementary and tend to reinforce each other. Rut the burden of proof seems
to be heavy for those who argue that politics involves a choice between these two dimensions of equality.
6. Summary and discussion
In this paper, I reviewed income distribution in Sweden according to
three dimensions of equality:
1. Cross-sectional equality of annual disposable income
2. Economic equality in a longer time perspective
3. Equality of opportunity as measured by correlations between income of fathers and sons
For obvious reasons, most empirical data are available for annual
data. Even though there are some conceptual and practical problems
involved in measuring disposable income over one year, there is now
quite good time-series information on this dimension of equality. It
seems, for example, as though Sweden has been able to retain its position among OECD countries as one of the four or five countries
with the most equal distributions of annual disposable income even
in the midst of the deep recession in the early 1990s.
It is sometimes argued that lifetime equality and equality of opportunity represent more basic and more valuable dimensions of equity. Of course this is partly a matter of values. But it is also an empirical question how much people are hurt by volatile income paths
or by temporarily low incomes. A few years of very low disposable
income for young students is probably not much of a problem,
whereas the consequences could be severe for families with children.
So it is important to learn more about the nature of the income variability that contributes to inequality of annual income but is neglected
in studies of lifetime income, before one can determine what weights
should be attached to these different dimensions of inequality.
Another important question is how different types of policy affect
these three different dimensions of equality. One hypothesis could be
that politics involves a choice between, on one hand, a society of
(say) the U.S. or the present UIC type with a high degree of crosssectional inequality but instead quite equal lifetime outcomes and high
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
degrees of equality of opportunity, and on the other hand a society of
the Swedish type with high cross-sectional equality but not lower
long-run equality or less equality of opportunity. An alternative hypothesis could be that these three dimensions of equality rather reinforce each other in the sense that smoothing of income paths also
contributes to more equal lifetime outcomes and more of equality of
opportunity.
My interpretation of the available evidence that I have reviewed in
this paper is that the latter hypothesis is the most plausible one. in
particular, three pieces of evidence speak in favor of this hypothesis
and against the first one. First, studies that have prolonged the time
period to longer periods than one single year show no sign of convergence in inequality between countries as the time unit is made
longer. Second, in Swedish data, inter-temporal income variability is
highest for persons with low long-run incomes, and taxes and transfers that smooth inter-temporal variability also equalize long-run income. Third, there are no indications that intergenerational connections are weaker in the U.S. and the UI(, rather the opposite pattern
can be found in data.
T o understand why this interpretation could be reasonable, it
might again be useful to consider families with children. If welfidre
state policies stabilize income paths for such fidmilies by sheltering
them from dips in market income, not only annual incomes are
equalized, but probably also lifetime incomes. A more stable private
economy for such families could also make them more prone to invest in the future of their children.
N o doubt, more research is needed to corroborate my somewhat
speculative conclusions. Some specific policies might also mainly affect annual incomes without any effects, or even regressive effects,
on more long-run outcomes. But if my conclusions are correct, it
becomes more important to tell politicians that the three dimensions
of equality are closely related to each other and that they tend to reinforce each other, rather than telling them that politics involves a
choice between these alternative dimensions of equality.
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
References
Aaberge, R., A. Bjorklund, M. Jiintti, M. Palme, P. Pedersen, N. Smith and T.
Wennemo (1996), Income Inequality and Income Mobility in the Scandinavian Countries Compared to the U.S., Discussion Papers No. 168 Statistics
Norway, Research Department.
Aaberge, R., A. Bjorklund, M. J h t t i , P. Pedersen, N. Smith and T. Wennemo
(1997, Unemployment Shocks and Income Distribution - How Did the
Nordic Countries Fare During their Crisis?, Discussion Papers No. 201
Statistics Norway, Research Department.
Altonji, J.G. and T.A. Dunn (1991), Relationships Among the Family Incomes and
Labor Market Outcomes of Relatives, Research in Labor Economics 12,
269-310.
Atkinson, A.B., L. Rainwater and T.M. Smeeding (1995), Income Distribution in
OECD countries: Evidence from the Luxembourg Income Study (LIS),
Paris: OECD.
Bjorklund, A. (1993), A Comparison Between Actual Distributions of *knual and
Lifetime Income: Sweden around 1970, Review of Income and Wealth, 39,
377-386.
Bjorklund, A. and M. Palme (1997, Income Redistribution within the Life Cycle
versus between Individuals: Empirical Evidence Using Swedish Panel Data,
Working Paper Series in Economics and Finance No. 197, Stockholm
School of Economics.
Bjorklund, A. and M. Jantti (1997a), Intergenerational Income Mobility in Sweden
Compared to the US., American Economic Review, 87, 1009-1018.
Bjorklund, A. and M. Jintti (1997b), Intergenerational Mobility of Economic
Status: Is the U.S. Different?, paper presented at the 1997 AEA-meeting in
New Orleans.
Bjorklund, A., M. Palme and I. Svensson (1995a), Assessing the Effects of Swedish
Tax and Benefit Reforms on Income Distribution using Different Income
Concepts, Tax Reform Evaluation Report No. 13, National Institute of
Economic Research, Stockholm.
Bjorklund, A., M. Palme and I. Svensson (1995b), Tax Reforms and Income Distribution: An Assessment Using Different Income Concepts, Swedish Economic Policy Review, 2, 229-265.
Blomquist, S. (1981), A Comparison of Distributions of Annual and Lifetime Income: Sweden around 1970, Review of Income and Wealth 27,243-264.
Burkhauser, R.V. and J.G. Poupore (1997, A Cross-National Comparison of Permanent Inequality in the U.S. and Germany, Review of Economics and
Statistics 79, 10-15.
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
Coder, J., L. Rainwater and T.M. Smeeding (1989), Inequality Among Children and
Elderly in Ten Modem Nations: The U.S. in an International Context,
American Economic Review 79,320-324.
Corak, M. and A. I-leisz (19961, Unto the Sons: The Intergenerational Income Mobility of Canadian Men, Statistics Canada.
Couch, 1C.A. and T.A. Dunn (1997), Intergenerational Correlations in Labor Market Status: A Comparison of the U.S. and Germany, Journal of Human Resources 32,210-232.
Dearden, L., S. Machin and H. Reed (1996), Intergenerational Mobility in Britain,
Economic Journal 107,47-66.
Edin, P.-A. and B. Holmlund (1995), The Swedish Wage Structure: The Rise and
Fall of Solidarity Wage Policy?, in R.B. Freeman and L.F. ICatz (eds.), Differences and Changes in Wage Structures. (Chicago: The University of Chicago Press).
Ellkson, R. and J. Goldthorpe (1992), The Constant Flux-a Study of Class Mobility in Industrial Sociebes (Clarendon, Oxford, England).
Fritzell, J. (1990), The Dynamics of Income Distribution: Economic Mobility in
Sweden in Comparison with the U.S., Social Science Research 19,17-46.
Fritzell, J. (1991), Icke av marknaden allena - Inkomstfordelningen i Sverige,
Swedish Institute for Social Research, Stockholm University.
Fritzell, J. (1992), OjBinlikhets- och fattigdomsutvecklmgen under 1980-talet:
Sverige i internationell belysning, Bilaga 8 till LU 92, Finansdepartementet
Stockholm.
Fritzell, J. (1994), Fordelningseffekter av offentliga tjhster, Rapport till ESO, Ds
1994:86, Finansdepartementet Stockholm.
Gottschalk, P. and T.M. Smeeding (1997, Cross-National Comparisons of Earnings and Income lnequali< ~ o u r n a lof Economic ~iterature,XXW, 633687.
Gustafsson, B. (1994), The Degree and Pattern of Income Mobility in Sweden,
Review of Income and Wealth 40, 67-86.
Gustafsson, B. and H. Uusitalo (1990), Income Distribution and Redistribution
during Two Decades: Experiences from Finland and Sweden, in I. Persson
(ed.), Generating Equality in the Welfare State - The Swedish Experience
(Norwegian University Press).
Gustafsson, B. and E . Palmer (1997), Changes in Swedish Inequality: A Studj- of
Equivalent Income, 1975-1991, in P. Gottschalk, B. Gustafsson and E.
Palmer (eds.), The Distribution of Economic Welfare in the 1980s
(Cambridge University Press).
Jansson, I<. (1990), Inkomst- och formogenhetsfordelnlngen 1967-1987, Bilaga 19
till LU90, Finansdepartementet, Stockholm.
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
Jenkins, S.P. and N.C. O'Leary (1996), Household Income Plus Household Production: The Distribubon of Extended Income in the UI<., Review of Income and Wealth 42,401-420.
Jintti, M. and S. Danziger (19941, Work, Welfare and Child Poverty in Sweden and
the US.,
Industrial and Labor Relations Review 48,48-64.
Jintti, M, and E. osterbacka (1995), How Much of the Variance in Income Can be
Attributed to Family Background? Empirical Evidence from Finland, Abo
Akademi TJniversity.
I<orpi, K. and J. Palme (1998, forthcoming), The Paradox of Redistribution and
Strategies of Equality: Welfare State Institutions, Inequality and Poverty in
the Western Countries, l h e r i c a n Sociological Review.
Lindbeck, A. (1997, The Swedish Experiment, Journal of Economic Literature
XXXV, 1273-1319.
Ministry of Finance (1996), Fordelningspolitisk redogorelse. Bilaga 4 till Prop.
1996/97:1.
Ministry of Finance (1997, Fordelningspolitisk redogorelse. Bilaga 7 till Prop.
1997/98:1.
OECD (1996), Employment Outlook (OECD, Paris).
OECD (1997), Employment Outlook (OECD, Paris).
Persson, M. and P.Wisskn (1984), Redistributional Aspects of Tax Evasion, Scandinavian Journal of Economics 86,131-149.
Rainwater, L,. and T.M. Smeeding (1995), Doing Poorly: The Real Income of
American Children in a Comparative Perspective, LIS-working paper series
No. 127.
Shorrocks, A. (1978), Income Inequality and Income Mobility, Journal of Economic Theory 19,376-393.
Smeeding, T.M., P. Saunders, J. Coder, S. Jenkins, J. Fritzell, A.J.M. Hagenaars, R.
Hauser, and M. Wolfson (1993), Poverty, Inequality, and Family Living
Standards Impacts Across Seven Nations: The Effect of Noncash Subsidies
for Health, Education and Housing, Review of Income and \T7ealth 39,
229-256.
Solon, G. (1992), Intergenerational Income llfob~l~ty
in the U.S., American Economic Review 82, 393-408.
Statistics Sweden (1997), Inkomstfordelningsundersakningen 1995.
Stihlberg, A.-C. (1990), Life cycle Income Redistribution of the Public Sector:
Inter- and Intra-generational Effects, in I. Persson (ed.), Generating Equality in the Welfare State: The Swedish Experience (Norwegan University
Press, Oslo).
INCOME DISTRIBUTION IN SWEDEN, Anders Bjorklund
Treiman, D.J. and H.B.G. Ganzeboom (199O), Cross-National Comparative StatusAttainment Research, Research in Social Straticification and Mobility 9,
105-127.
Uddhammar, E. (1997), Arbete, vdfard, bidrag - en dynamisk analys av folkets
v d s t h d och valfards forskningens miss fijrsthd (City University Press,
Stockholm).
Wiegand, J. (1997), Intergenerational Earnings Mobility in Germany, mimeo University College London.
Zandvaluh, A. and B. Gusrafsson (1997,Dynamics of income i n e q u a l i ~in Sweden, paper presented at the 1997-meeting of the European Society of
Population Economics.
Zirnmerman, D.J. (1992), Regression Toward Mediocrity in Economic Stature,
American Economic Review 82,409-429.
Aberg, R. (1989), Distributive Mechanisms of the Welfare State - a formal analysis
and an empirical application, European Sociological Review 5, 167-182.